Split out constant portion of PhraseSplitter to PhraseSplitterQueryEnv.

5 files changed, 169 insertions, 107 deletions

diff --git a/searchlib/src/vespa/searchlib/fef/CMakeLists.txt b/searchlib/src/vespa/searchlib/fef/CMakeLists.txt
index 08e64701c05..396775b20c5 100644
--- a/searchlib/src/vespa/searchlib/fef/CMakeLists.txt
+++ b/searchlib/src/vespa/searchlib/fef/CMakeLists.txt
@@ -23,6 +23,7 @@ vespa_add_library(searchlib_fef OBJECT
     parameter.cpp
     parameterdescriptions.cpp
     parametervalidator.cpp
+    phrase_splitter_query_env.cpp
     phrasesplitter.cpp
     properties.cpp
     queryproperties.cpp
diff --git a/searchlib/src/vespa/searchlib/fef/phrase_splitter_query_env.cpp b/searchlib/src/vespa/searchlib/fef/phrase_splitter_query_env.cpp
new file mode 100644
index 00000000000..b474317a33b
--- /dev/null
+++ b/searchlib/src/vespa/searchlib/fef/phrase_splitter_query_env.cpp
@@ -0,0 +1,76 @@
+// Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+#include "phrase_splitter_query_env.h"
+
+namespace search::fef {
+
+void
+PhraseSplitterQueryEnv::considerTerm(uint32_t termIdx, const ITermData &term, std::vector<PhraseTerm> &phraseTerms, uint32_t fieldId)
+{
+    typedef search::fef::ITermFieldRangeAdapter FRA;
+
+    for (FRA iter(term); iter.valid(); iter.next()) {
+        if (iter.get().getFieldId() == fieldId) {
+            TermFieldHandle h = iter.get().getHandle();
+            _maxHandle = std::max(_maxHandle, h);
+            if (term.getPhraseLength() > 1) {
+                SimpleTermData prototype;
+                prototype.setWeight(term.getWeight());
+                prototype.setPhraseLength(1);
+                prototype.setUniqueId(term.getUniqueId());
+                prototype.addField(fieldId);
+                phraseTerms.push_back(PhraseTerm(term, _terms.size(), h));
+                for (uint32_t i = 0; i < term.getPhraseLength(); ++i) {
+                    _terms.push_back(prototype);
+                    _termIdxMap.push_back(TermIdx(_terms.size() - 1, true));
+                }
+                return;
+            }
+        }
+    }
+    _termIdxMap.push_back(TermIdx(termIdx, false));
+}
+
+PhraseSplitterQueryEnv::PhraseSplitterQueryEnv(const IQueryEnvironment & queryEnv, uint32_t fieldId) :
+    _queryEnv(queryEnv),
+    _terms(),
+    _termIdxMap(),
+    _maxHandle(0),
+    _skipHandles(0)
+{
+    TermFieldHandle numHandles = 0; // how many handles existed in underlying data
+    std::vector<PhraseTerm> phraseTerms; // data about original phrase terms
+
+    for (uint32_t i = 0; i < queryEnv.getNumTerms(); ++i) {
+        const ITermData *td = queryEnv.getTerm(i);
+        assert(td != nullptr);
+        considerTerm(i, *td, phraseTerms, fieldId);
+        numHandles += td->numFields();
+    }
+
+    _skipHandles = _maxHandle + 1 + numHandles;
+    TermFieldHandle term_handle = _skipHandles;
+    for (auto & term : _terms) {
+        // start at _skipHandles + 0
+        term.field(0).setHandle(term_handle);
+        ++term_handle;
+    }
+
+    for (uint32_t i = 0; i < phraseTerms.size(); ++i) {
+        const PhraseTerm &pterm = phraseTerms[i];
+
+        for (uint32_t j = 0; j < pterm.term.getPhraseLength(); ++j) {
+            const ITermData &splitp_td = _terms[pterm.idx + j];
+            const ITermFieldData& splitp_tfd = splitp_td.field(0);
+            HowToCopy meta;
+            meta.orig_handle = pterm.orig_handle;
+            meta.split_handle = splitp_tfd.getHandle();
+            meta.offsetInPhrase = j;
+            _copyInfo.push_back(meta);
+        }
+    }
+}
+
+PhraseSplitterQueryEnv::~PhraseSplitterQueryEnv() = default;
+
+}
diff --git a/searchlib/src/vespa/searchlib/fef/phrase_splitter_query_env.h b/searchlib/src/vespa/searchlib/fef/phrase_splitter_query_env.h
new file mode 100644
index 00000000000..19a4cb206b9
--- /dev/null
+++ b/searchlib/src/vespa/searchlib/fef/phrase_splitter_query_env.h
@@ -0,0 +1,85 @@
+// Copyright Verizon Media. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
+
+#pragma once
+
+#include "iqueryenvironment.h"
+#include "simpletermdata.h"
+
+namespace search::fef {
+
+/**
+ * This class is used to split all phrase terms in a query environment
+ * into separate terms. New TermData and TermFieldMatchData objects
+ * are created for each splitted phrase term and managed by this
+ * class.  Unmodified single terms are served from the query
+ * environment and match data.
+ *
+ * The TermFieldMatchData objects managed by this class are updated
+ * based on the TermFieldMatchData objects associated with the
+ * original phrase terms. Positions are adjusted with +1 for each term
+ * after the first one.
+ *
+ * Use this class if you want to handle a phrase term the same way as
+ * single terms.
+ **/
+class PhraseSplitterQueryEnv : public IQueryEnvironment
+{
+protected:
+    struct TermIdx {
+        uint32_t idx;      // index into either query environment or vector of TermData objects
+        bool     splitted; // whether this term has been splitted or not
+        TermIdx(uint32_t i, bool s) : idx(i), splitted(s) {}
+    };
+    struct PhraseTerm {
+        const ITermData & term; // for original phrase
+        uint32_t idx; // index into vector of our TermData objects
+        TermFieldHandle orig_handle;
+        PhraseTerm(const ITermData & t, uint32_t i, uint32_t h) : term(t), idx(i), orig_handle(h) {}
+    };
+    struct HowToCopy {
+        TermFieldHandle orig_handle;
+        TermFieldHandle split_handle;
+        uint32_t offsetInPhrase;
+    };
+
+    const IQueryEnvironment        &_queryEnv;
+    std::vector<SimpleTermData>     _terms;       // splitted terms
+    std::vector<HowToCopy>          _copyInfo;
+    std::vector<TermIdx>            _termIdxMap;  // renumbering of terms
+    TermFieldHandle                 _maxHandle;   // the largest among original term field handles
+    TermFieldHandle                 _skipHandles;   // how many handles to skip
+
+    void considerTerm(uint32_t termIdx, const ITermData &term, std::vector<PhraseTerm> &phraseTerms, uint32_t fieldId);
+
+public:
+    /**
+     * Create a phrase splitter based on the given query environment.
+     *
+     * @param queryEnv the query environment to wrap.
+     * @param field the field where we need to split phrases
+     **/
+    PhraseSplitterQueryEnv(const IQueryEnvironment & queryEnv, uint32_t fieldId);
+    ~PhraseSplitterQueryEnv();
+
+    /**
+     * Update the underlying TermFieldMatchData objects based on the bound MatchData object.
+     **/
+    uint32_t getNumTerms() const override { return _termIdxMap.size(); }
+
+    const ITermData * getTerm(uint32_t idx) const override {
+        if (idx >= _termIdxMap.size()) {
+            return nullptr;
+        }
+        const TermIdx & ti = _termIdxMap[idx];
+        return ti.splitted ? &_terms[ti.idx] : _queryEnv.getTerm(ti.idx);
+    }
+
+    const Properties & getProperties() const override { return _queryEnv.getProperties(); }
+    const Location & getLocation() const override { return _queryEnv.getLocation(); }
+    const attribute::IAttributeContext & getAttributeContext() const override { return _queryEnv.getAttributeContext(); }
+    double get_average_field_length(const vespalib::string &field_name) const override { return _queryEnv.get_average_field_length(field_name); }
+    const IIndexEnvironment & getIndexEnvironment() const override { return _queryEnv.getIndexEnvironment(); }
+};
+
+
+}
diff --git a/searchlib/src/vespa/searchlib/fef/phrasesplitter.cpp b/searchlib/src/vespa/searchlib/fef/phrasesplitter.cpp
index e84f61332e1..a0d53b12fe5 100644
--- a/searchlib/src/vespa/searchlib/fef/phrasesplitter.cpp
+++ b/searchlib/src/vespa/searchlib/fef/phrasesplitter.cpp
@@ -4,74 +4,16 @@
 
 namespace search::fef {
 
-void
-PhraseSplitter::considerTerm(uint32_t termIdx, const ITermData &term, std::vector<PhraseTerm> &phraseTerms, uint32_t fieldId)
-{
-    typedef search::fef::ITermFieldRangeAdapter FRA;
-
-    for (FRA iter(term); iter.valid(); iter.next()) {
-        if (iter.get().getFieldId() == fieldId) {
-            TermFieldHandle h = iter.get().getHandle();
-            _maxHandle = std::max(_maxHandle, h);
-            if (term.getPhraseLength() > 1) {
-                SimpleTermData prototype;
-                prototype.setWeight(term.getWeight());
-                prototype.setPhraseLength(1);
-                prototype.setUniqueId(term.getUniqueId());
-                prototype.addField(fieldId);
-                phraseTerms.push_back(PhraseTerm(term, _terms.size(), h));
-                for (uint32_t i = 0; i < term.getPhraseLength(); ++i) {
-                    _terms.push_back(prototype);
-                    _termIdxMap.push_back(TermIdx(_terms.size() - 1, true));
-                }
-                return;
-            }
-        }
-    }
-    _termIdxMap.push_back(TermIdx(termIdx, false));
-}
-
-PhraseSplitter::PhraseSplitter(const IQueryEnvironment & queryEnv, uint32_t fieldId) :
-    _queryEnv(queryEnv),
-    _matchData(nullptr),
-    _terms(),
-    _termMatches(),
-    _termIdxMap(),
-    _maxHandle(0),
-    _skipHandles(0)
+PhraseSplitter::PhraseSplitter(const IQueryEnvironment & queryEnv, uint32_t fieldId)
+    : PhraseSplitterQueryEnv(queryEnv, fieldId),
+      _matchData(nullptr),
+      _termMatches()
 {
-    TermFieldHandle numHandles = 0; // how many handles existed in underlying data
-    std::vector<PhraseTerm> phraseTerms; // data about original phrase terms
-
-    for (uint32_t i = 0; i < queryEnv.getNumTerms(); ++i) {
-        const ITermData *td = queryEnv.getTerm(i);
-        assert(td != nullptr);
-        considerTerm(i, *td, phraseTerms, fieldId);
-        numHandles += td->numFields();
-    }
-
-    _skipHandles = _maxHandle + 1 + numHandles;
     _termMatches.reserve(_terms.size());
-    for (auto & term : _terms) {
-        // start at _skipHandles + 0
-        term.field(0).setHandle(_skipHandles + _termMatches.size());
+    for ([[maybe_unused]] auto & term : _terms) {
         _termMatches.emplace_back();
         _termMatches.back().setFieldId(fieldId);
     }
-
-    for (uint32_t i = 0; i < phraseTerms.size(); ++i) {
-        const PhraseTerm &pterm = phraseTerms[i];
-
-        for (uint32_t j = 0; j < pterm.term.getPhraseLength(); ++j) {
-            const ITermData &splitp_td = _terms[pterm.idx + j];
-            const ITermFieldData& splitp_tfd = splitp_td.field(0);
-            HowToCopy meta;
-            meta.orig_handle = pterm.orig_handle;
-            meta.split_handle = splitp_tfd.getHandle();
-            meta.offsetInPhrase = j;
-            _copyInfo.push_back(meta);
-        }
-    }
 }
 
 PhraseSplitter::~PhraseSplitter() = default;
diff --git a/searchlib/src/vespa/searchlib/fef/phrasesplitter.h b/searchlib/src/vespa/searchlib/fef/phrasesplitter.h
index 8b399885496..dc7954a7fcc 100644
--- a/searchlib/src/vespa/searchlib/fef/phrasesplitter.h
+++ b/searchlib/src/vespa/searchlib/fef/phrasesplitter.h
@@ -2,11 +2,9 @@
 
 #pragma once
 
-#include "iqueryenvironment.h"
+#include "phrase_splitter_query_env.h"
 #include "matchdata.h"
-#include "simpletermdata.h"
 #include "termfieldmatchdata.h"
-#include "fieldinfo.h"
 
 namespace search::fef {
 
@@ -25,36 +23,10 @@ namespace search::fef {
  * Use this class if you want to handle a phrase term the same way as
  * single terms.
  **/
-class PhraseSplitter : public IQueryEnvironment
+class PhraseSplitter : public PhraseSplitterQueryEnv
 {
-private:
-    struct TermIdx {
-        uint32_t idx;      // index into either query environment or vector of TermData objects
-        bool     splitted; // whether this term has been splitted or not
-        TermIdx(uint32_t i, bool s) : idx(i), splitted(s) {}
-    };
-    struct PhraseTerm {
-        const ITermData & term; // for original phrase
-        uint32_t idx; // index into vector of our TermData objects
-        TermFieldHandle orig_handle;
-        PhraseTerm(const ITermData & t, uint32_t i, uint32_t h) : term(t), idx(i), orig_handle(h) {}
-    };
-    struct HowToCopy {
-        TermFieldHandle orig_handle;
-        TermFieldHandle split_handle;
-        uint32_t offsetInPhrase;
-    };
-
-    const IQueryEnvironment        &_queryEnv;
     const MatchData                *_matchData;
-    std::vector<SimpleTermData>     _terms;       // splitted terms
     std::vector<TermFieldMatchData> _termMatches; // match objects associated with splitted terms
-    std::vector<HowToCopy>          _copyInfo;
-    std::vector<TermIdx>            _termIdxMap;  // renumbering of terms
-    TermFieldHandle                 _maxHandle;   // the largest among original term field handles
-    TermFieldHandle                 _skipHandles;   // how many handles to skip
-
-    void considerTerm(uint32_t termIdx, const ITermData &term, std::vector<PhraseTerm> &phraseTerms, uint32_t fieldId);
 
     TermFieldMatchData *resolveSplittedTermField(TermFieldHandle handle) {
         return &_termMatches[handle - _skipHandles];
@@ -88,15 +60,6 @@ public:
      * Update the underlying TermFieldMatchData objects based on the bound MatchData object.
      **/
     void update();
-    uint32_t getNumTerms() const override { return _termIdxMap.size(); }
-
-    const ITermData * getTerm(uint32_t idx) const override {
-        if (idx >= _termIdxMap.size()) {
-            return nullptr;
-        }
-        const TermIdx & ti = _termIdxMap[idx];
-        return ti.splitted ? &_terms[ti.idx] : _queryEnv.getTerm(ti.idx);
-    }
 
     /**
      * Inherit doc from MatchData.
@@ -108,11 +71,6 @@ public:
         return handle < _skipHandles ? _matchData->resolveTermField(handle) : resolveSplittedTermField(handle);
     }
 
-    const Properties & getProperties() const override { return _queryEnv.getProperties(); }
-    const Location & getLocation() const override { return _queryEnv.getLocation(); }
-    const attribute::IAttributeContext & getAttributeContext() const override { return _queryEnv.getAttributeContext(); }
-    double get_average_field_length(const vespalib::string &field_name) const override { return _queryEnv.get_average_field_length(field_name); }
-    const IIndexEnvironment & getIndexEnvironment() const override { return _queryEnv.getIndexEnvironment(); }
     void bind_match_data(const fef::MatchData &md) { _matchData = &md; }
 };